key: cord-0262972-lpxaka1b
authors: Tilley, A.; Byrd, K. A.; Pincus, L.; Klumpyan, K.; Dobson, K.; Dos Reis Lopes, J.; Shikuku, K.
title: A randomised controlled trial to test the effects of fish aggregating devices and SBC activities promoting fish consumption in Timor-Leste: A study protocol
date: 2021-08-12
journal: nan
DOI: 10.1101/2021.08.10.21261568
sha: ff51a5809fd30ace7a2d296602b929fe319999b5
doc_id: 262972
cord_uid: lpxaka1b

Timor-Leste is one of the world's most malnourished nations where micronutrient-deficient diets are a contributing factor to the prevalence of child stunting, currently estimated to be 45.6% of children under five. Fish are an important source of micronutrients and one that may assist the country's predominantly rural population of agriculturalists to exit poverty and malnutrition. However, a small national fishing fleet producing low catch volumes place fish out of reach of most inland and upland populations where it is needed most. Fish consumption is very low in rural areas compared to coastal, regional, and global averages. This trial is a one-year, cluster-randomized, partially masked, controlled trial among families living in rural, inland Timor-Leste. This trial aims to test and compare the effects of two treatments, alone and in combination, on the frequency and volume of household fish consumption in upland areas as a proxy for improved dietary diversity and micronutrient intake. Treatment 1 is the installation of nearshore, moored fish aggregating devices (FADs) to improve catch rates with existing fishing gears. Treatment 2 is social and behaviour change (SBC) activities to promote fish consumption. Villages in inland communities will be randomized to receive treatment 1, treatment 2, both treatments, or neither treatment. Some households with one child under five will be recruited, and data will be collected at baseline (prior to the rollout of the treatments) and endline. Our study will determine the impact of an improved supply of fish, along with nutrition-oriented SBC activities, on the fish purchasing and consumption practices of rural, inland households. Findings from this study are urgently needed by small island developing states in order to make policy and investment decisions on how best to improve households' diets using locally available, nutrient-dense foods such as fish. Investments such as these are needed to break the cycle of malnutrition.

Timor-Leste is a country afflicted by multiple burdens of malnutrition, and sustains one of the 33

world's highest rates of stunting (low height for age), afflicting 47.1% of children under five [1] . 34

Stunted growth is multifactorial, but in Timor-Leste, a contributing cause is likely to be a 35 monotonous diet [2] . There is mounting evidence that fish is an underutilised resource that has 36 the potential to improve dietary quality, but consumption of fish is low at ~6 kg/person/yr [3] 37 compared to neighbouring countries in Asia and the global average of ~20.5 kg/person/yr [4] . 38

However, this 6 kg average figure masks the large variation in fish consumption; coastal 39 communities consume ~17 kg/person/year, while inland communities consume fish only 40 seasonally, resulting in a consumption rate of ~4 kg/person/yr [3] . 41

Fish consumption in inland and rural regions of Timor-Leste is constrained in part due to the low 42 catch volumes landed by the small national fleet of mostly non-motorized canoes [5] . Additional 43 barriers, such as limited road and market infrastructures in inland areas, also reduce inland 44

consumers' access to fish. Decades of low access to marine fish in inland areas [3] has decreased 45 consumers' expectations of consuming this food source and made them unlikely to demand it 46 from local traders. 47

Fish aggregating devices (FADs) are a technology to concentrate fish to make them easier to find 48 and catch, and there is evidence of their improving catch rates in both inshore [6, 7] and offshore 49 settings [8, 9] . Offshore FADs can be anchored or drifting, and are used to target tuna and other 50

highly valued pelagic fish [10, 11] . Anchored nearshore FADs are utilized by shore-based fishers 51

to catch small pelagic fishes such as mackerels and scads [6] . By enabling access to a previously 52 inaccessible fish stock, FADs can increase the climate adaptive capacity of fishing communities 53 by diversifying the species caught and by reducing the dependence on production from vulnerable 54 reef habitat [13] . 55

Investigating the barriers for fish from small-scale fishers to reach the populations in need is 56 important for understanding the contribution of fisheries to nutrition security. Proximity to inland 57 fisheries is associated with an increase in fish consumption in children in rural Sub-Saharan Africa 58

[14]; however, more information is needed to determine if this relationship holds in other contexts. 59

Fish aggregating devices can increase supply, but increased availability alone may not be 60 adequate to increase consumption, especially when it comes to using animal-source foods for 61

infant and child feeding [15] . Building consumers' awareness and demand for high-quality marine 62

fish must go hand-in-hand with increasing fish supply to improve the contribution of fisheries to 63 nutrition security in inland areas. 64

Beyond increasing supply and awareness on benefits of fish consumption, a woman's decision-65 making power, sociocultural practices and seasonality have strong impacts on a household's 66 ability to consume fish. A Gender Equality and Social Inclusion Analysis (GESIA) conducted in 67

2017 [15, 16] found that a major gender equity issue in nutrition relates to the constraints that 68

women face in decision-making and their lack of control over food, which impacts significantly on 69 nutrition choices and the prevention of malnutrition. Research has also found that mothers 70 (younger mothers in particular) were not confident to spend money on small quantities of meat or 71 canned fish without permission from their husband [17, 18] . 72

Promoting a healthy diet and optimal infant and child feeding practices through contextualized 73 SBC interventions is another avenue to improve dietary quality and eventual health outcomes. 74

The 

Our trial aims to evaluate the effectiveness of increased fish production and targeted promotion 92 of the nutritional benefits of fish consumption through social behaviour change (SBC) activities on 93 the frequency and quantity of fish consumption in rural households in Timor-Leste. Findings from 94 our study will inform governments, practitioners, and donors on the implementation of effective 95

interventions to improve household diets through increased fish consumption. 96

H0 Deployment of FADs has no effect on volume and frequency of fish consumed 98 H0 Providing SBC has no effect on volume and frequency of fish consumed 99 H0 Combined deployment of FADs and SBC has no effect on volume and frequency of fish 100 consumed 101 H0 Deployment of FADs and SBC have no differential impacts on consumption 102 2.3 Trial design 103 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted August 12, 2021. ; https://doi.org/10.1101/2021.08.10.21261568 doi: medRxiv preprint Our study will implement a one year-long, cluster randomised controlled trial (RCT) among rural 104 inland households. We will test the effects of a bundled intervention addressing both supply-side 105

and demand-side constraints to consumption of fish.3Intervention 106

On the supply side, the experiment will involve the deployment of two fish aggregating devices in 107 the nearshore fishing grounds of a coastal fishing village with already established trading links to 108

inland study sites. Randomisation of the interventions will occur at the site level. Using WorldFish 109

and government data on fisheries landing sites in six municipalities, rural coastal fishing villages 110

with over 10 active fishing vessels will be listed. Villages that already have or utilise FADs will be 111 excluded. Six villages (one per municipality) will then be randomly assigned to one of two RCT 112 arms, namely FAD (4 villages) and non-FAD (2 villages) ( Figure 2 ). This treatment arm is 113

imbalanced 2:1 because although the efficacy of nearshore FADs at increasing catch rates of fish 114 has been established [6], it is also shown to be dependent upon local ecological and bathymetric 115

conditions. These coastal villages are geographically dispersed, hence minimizing the risk of 116 contamination between treatment and control sites. In addition, by mapping the value chains of 117 traders in baseline focus groups, we will minimise contamination of non-FAD sites with fish caught 118 at a FAD. 119

On the demand side, 3-6 inland villages within a 30 km radius of the coastal treatment sites that 120 have a VSLA established by Mercy Corps, will be further randomised into an SBC arm (treatment) 121

and a non-SBC arm (control). A further inclusion criterion for the villages is that fish traders from 122 the coastal landing sites confirm through interview that they currently sell fish products to those 123 respective villages. Villages will be randomly allocated to treatment or control arms using an Excel 124 random number table to a 1:1 allocation ratio. The principal researcher will generate the allocation 125 sequence. 126

Within each SBC inland village, Mercy Corps will layer nutrition content and learning through 127

VSLAs. The VSLA members will participate in skill-building activities that promote five key 128 behaviours 1 about fish nutrition and household decision making on fish consumption as part of 129 facilitated discussions, interactive learning sessions, and a video-based facilitated dialogue. At 130 the village level (N = 12), Mercy Corps will promote increased fish consumption through SBC 131 mass attendance events through marketing of fish vendors, interactive learning, and edu-tainment 132

competitions promoting fish consumption. SBC will raise awareness of the benefits of eating fish 133 as part of a balanced diet, with a particular focus on pregnant and lactating mothers, and children 134

under 5 years of age. The non-SBC communities will not receive the nutrition campaign package. 135 1 Key promoted behaviours include: 1) Incorporate fish into family meals at least twice a week; 2) Parents pick bones out of fish for small children and start offering fish to infants at 6 months of age; 3) VSLA members save with a purpose of protein consumption for their families, with a focus on fish consumption; 4) Couples initiate conversations on a weekly basis on allocation of resources for weekly protein purchase, inclusive of fish; 5) Households use resources to purchase fish two times a week.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted August 12, 2021. ; https://doi.org/10.1101/2021.08.10.21261568 doi: medRxiv preprint This study will be carried out in six municipalities of Timor-Leste (Bobonaro, Dili, Liquica, 141

Manufahi, Manatuto and Covalima) ( Figure 3 ). The rural, upland areas of these municipalities 142 exhibit some of the highest rates of child stunting in the world at 47.1% [26]. 143 The sampling frame comprises rural, inland households in Timor Leste. We conducted a power 147 calculation using the command clustersampsi in STATA based on secondary data, and a primary 148 outcome of fish consumption per household. We assumed a minimum detectable effect of 15% 149

and an intra-cluster correlation coefficient equal to 0.05. In order to achieve at least 80% statistical 150 power, we require a sample size of 575 inland community households selected from 25 151 communities (that is, we sample 23 households in each community). Adjusting for 25% attrition, 152

we estimate that a sample size of 700 is required-about 30 households per site. 153

Sampling follows a multi-stage procedure. In the first stage, 6 coastal municipalities were 154

purposively selected based on the coverage of Mercy Corps programming of VSLA groups. The 155 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted August 12, 2021. ; https://doi.org/10.1101/2021.08.10.21261568 doi: medRxiv preprint second step involves compiling a list of rural coastal fishing villages with over 10 active fishing 156 vessels and randomly selecting one village within each municipality. In the third step, a list of 157

inland villages within a 30 km radius of each coastal villages and with active presence of a VSLA 158 was compiled. Although we include only villages participating in VSLAs, the typical size of a VSLA 159

(about 15 members) is less than the required number of households per village. In the fourth step, 160 therefore, we generate a list of all households in the village, both VSLA members and non-161 members. All VSLA participating households are selected to participate in the study. In order to 162 ensure the required sample size is satisfied (in accordance with the power analysis) we randomly 163 select additional households from the list of those not participating in VSLAs. The chosen 164

households will be visited and asked if they would be willing to participate in the study. Where 165

households are not willing, the closest neighbouring household will be asked, and so on. 166

Communities and study households will be informed of the study and sensitized by the research 168 team to its activities and purpose through community meetings and meetings with local leaders. 169

Treatment and control groups will not be informed of their assigned group to avoid the possibility 170 of introducing Hawthorne or John Henry effects. An informed consent including maintenance of 171 the confidentiality of personal data, and the possibility to refuse the consent without having to 172 justify the refusal, will be obtained from all community leaders and study participants prior to 173 responding to any surveys (see Appendix A). 174 2.7 Data collection 175

At each of the six coastal sites, a baseline of fishing activities will be collected from participating 177 fishers including trip frequency, duration, method, location and substrate type, along with the 178 catch volume. Data will be collected from fishers as they return to shore by on-site enumerators, 179 consistent with the national fisheries monitoring system in Timor-Leste, PeskAAS (following [24] ). 180

This is a tablet-based survey where government enumerators collect and upload near-real-time 181

landings data to the central database. All participation in vessel tracking and contribution of data 182 is voluntary. 183

Information collected includes the number of fishers (sex-disaggregated), boat and gear types, 184

individuals sizes and quantities of fish captured, trip duration, the primary habitat fished, and the 185 proportion and price of fish sold. These data are also paired with geospatial vessel tracks for 186

boats that are fitted with a GPS tracker. Fishing activities and catch monitoring will be established 187

prior to installation of any FADs, to evaluate any change in catch rates or production volume 188

following the FAD deployment. Enumerators at each of the six coastal sites will be responsible 189

for daily data collection at landing sites of small-scale fishers, and will collect data on fish catch 190 by family group. 191

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted August 12, 2021. ; https://doi.org/10.1101/2021.08.10.21261568 doi: medRxiv preprint Following enrolment and informed consent, a baseline survey will be conducted with one member 193 per household. The baseline survey will consist of household demographic information, asset 194 ownership, and food consumption questions to assess relative wealth and standard of living, 195

access to fish and other animal source foods, food insecurity, and the frequency and volume of 196 fish consumed at the household level. Qualitative 24-hour recalls will be done with the household 197 member responsible for food preparation. If there are children under five years old in the 198 household and the respondent is not the primary caregiver, the caregiver will be asked 24-hour 199 recall questions related to the diet of the child(ren). The 24-hour recalls will be used to calculate 200 the minimum dietary diversity for both the woman and child [25, 26] . We will also ascertain if there 201 are any cultural or traditional beliefs that influence fish consumption for specific people or at 202 certain times. 203

Interviews will be conducted by enumerators extensively trained on the tools, methods, research 204 ethics and confidentiality. Questionnaires will be translated into Tetum and carried out in Tetum 205 or a combination of Tetum and other local languages. After enumerator training and survey pre-206

testing, the survey tools and procedures will be adapted as necessary to account for errors in 207

interpretation by enumerators and respondents. Staff will inform the village chief and VSLA 208 members in advance of the baseline and endline data collection sessions, and the latter will 209

arrange for participants to be present. 210 2.7 Data management 211 All data will be kept confidential. Data will be submitted daily via 3G for review by in-country 212 researchers and only stored on password protected mobile phone tablets and laptops temporarily.

The household survey questionnaire will include skip logics and field specifications and delimiters 214 to minimise input errors, but also supervising researchers will flag any unusual data for checking 215 with enumerators. All variables will be visualized and checked for outliers. Outliers will be closely 216

inspected, and sensitivity analyses will be conducted on high leverage data points. Outcome 217 variables will be visually checked for normality and transformed if necessary. Participants will be 218 assigned randomly-generated ID numbers, and names and identifying information will be de-219

linked from the dataset. Data will be analysed for groups and no individual response will be 220 identifiable. Survey data will be archived and accessible to only the research team and those 221 approved by the entire research team. Anonymised datasets will be securely maintained 222

indefinitely to facilitate future research for example pooled analysis with other studies for purposes 223 of meta-analysis or multi-site analysis. 224 3) Average catch rates calculated as the increase in fish availability by comparing average 238 catch rates prior to FAD deployment with catch rates post-deployment. Catch rate and 239

total production values for different coastal community sites will be calculated and 240 extrapolated following Tilley et al. [24] . 

assessing knowledge of nutritional benefits of fish and handling practices will be 246 administered to respondents at baseline, midline and endline. 247

The main treatment variable is a dummy variable that indicates to which of the four treatment 249 arms a village was assigned. Our primary model is an unadjusted model. 250

However, if we find that after unmasking, there are variables in the dataset that appear different 251 by treatment arm, we may consider adjusting for confounders in the model. 252 253 We will assess whether the effect of SBC is mediated by an exogenous exposure of coastal 254 communities to the FADs. Therefore, we will construct an interaction of the SBC treatment dummy 255 with a dummy variable equal to one if an inland community is linked to a coastal community with 256 access to a FAD and zero otherwise. 257

Additionally, we will assess whether increased fish consumption is mediated by a change in self-258 reported knowledge. We will do this by constructing a dummy composite variable of all knowledge 259 questions, and looking for an interaction by knowledge and treatment arm on per capita fish 260

intakes. 261 2.10 Checking for spillover effects 262 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted August 12, 2021. groups. However, we cannot rule out that attrition may still occur. Therefore, in case of substantial 275 attrition, the reasons will be carefully recorded. This will help us to assess whether potential 276 causes of attrition are systematically linked to specific treatments. We will check whether attrition 277

rates are equal across the three experimental arms or if concentrated in one arm. High attrition is 278 potentially problematic, as it could introduce selection bias in our randomised experiment. In case 279

of high attrition, we will examine the implication of attrition for our results in several ways. First, 280

we will test whether attrition is affected by treatment assignment. If this is the case, we will 281

implement Lee bounds to estimate lower and upper bounds of treatment effects [28, 29] . Second, 282

we will test whether our remaining sample is (still) balanced along key observable dimensions, 283

and whether attrition is driven by different characteristics across treatment groups. To do this, a 284 probit regression will be used where we regress attrition status on the treatment dummies, 285

household characteristics, and the interaction of these. The hypothesis that attrition is not 286 differentially determined by treatment status will be assessed through an F-test for the joint 287 significance of the interaction terms in this regression. If differential determinants of attrition are 288 observed, we will attempt to control for this through a weighting procedure as a robustness 289

analysis [30] . Specifically, we will follow a two-stage procedure. In the first stage, a logit regression 290 is used to estimate the predicted probability of having non-missing measures for our outcomes 291

given treatment assignment and a vector of observable covariates. In the second stage, we will 292 weight each observation using the inverse of the thus estimated probability of having a non-293 missing measure of our outcomes. We will check whether our main results remain robust to all 294 these robustness tests. 295 2.12 Multiple hypothesis testing 296 Because we are making inference on a large number of hypotheses, it is possible that significant 297 results emerge from our analysis due to chance rather than actual treatment effects. We will follow 298

[31] and adjust the p-values using a number of different methods. We will calculate Romano-Wolf 299

adjusted p-values (following [32]) to correct for the familywise error rate (FWER), the probability 300 of making at least one false discovery among a family of comparisons. We will also calculate 301

sharpened q-values (as in [33] ) to correct for the false discovery rate (FDR), the probability of 302 making at least one false discovery among the discoveries already made. 303 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted August 12, 2021. ; https://doi.org/10.1101/2021.08.10.21261568 doi: medRxiv preprint 2.13 Balance at baseline 304 The goal of the randomization is to ensure that the treatment and control groups are similar in 305 terms of average characteristics at baseline. This will be tested formally through an F-test of joint 306 orthogonality using a multinomial logit regression, which tests whether the observable 307 characteristics in 

Intent-to-treat (ITT) estimates will be presented following equation 1: 316

where is the outcome variable for household in village (Per capita household fish 318 consumption, frequency of fish consumed at household level, frequency of fish consumption in 319

women, frequency of fish consumption in children, knowledge about nutrition benefits of fish). The 320 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted August 12, 2021. ; https://doi.org/10.1101/2021.08.10.21261568 doi: medRxiv preprint variable denotes the treatment dummy variables. is a vector of control variables 321

including household characteristics and baseline levels of outcome variables, captures 322 municipality fixed effects, and is the usual idiosyncratic error term, clustered at the village 323 level (the unit of randomization). Here the parameter of interest is , the average treatment effect. 324

Equation will be estimated using OLS. We will first estimate a parsimonious model with only the 325 treatment dummies and the municipality fixed effects. Then we will control for additional baseline 326

covariates. The control group in equation 1 is the no-FAD and no-SBC group. Equation 1 will be 327 estimated separately for midline and endline subsamples. However, we will compare the 328 magnitude of impacts at midline and endline. 329 2.14 Ethical and safety considerations FADs effect on catch rates was highly dependent on site ecology, but on average FADs provided 355 a return on investment in five months or less. Furthermore, the oceanographic characteristics of 356 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted August 12, 2021. ; https://doi.org/10.1101/2021.08.10.21261568 doi: medRxiv preprint Timor-Leste (steep slopes and strong currents) limit the longevity of deployed FADs, so one year 357

is sufficient time to see any changes in catch rates. 358

Our study seeks to understand if increased fish volumes in coastal areas is absorbed by current 359 markets, or is sufficient surplus to drive greater supply into inland areas. It is designed to approach 360 the push and pull market dynamic from either end, without directly incentivising actors in the 361 market chain. By increasing the volume of fish available for sale, we are pushing surplus supply 362

into new and smaller markets; and by encouraging fish consumption through a contextualized 363 SBC intervention, we aim to increase the demand for fish to pull more into rural market. We do 364

not anticipate large effects on the market price of fish in the study timeline, but we will record 365 fisher sale price as part of monitoring. The reason for not incentivising traders to go to specific 366 study villages is the lack of a reliable mechanism through which this could be sustained beyond 367

the project life cycle. Individual variation between traders and their extant connections with inland 368 communities is likely to be an important factor in decisions over where to direct greater fish 369 production. We aim to mitigate this as far as possible through selection of traders for baseline and 370 endline focus groups that already trade in the study inland communities. 371

While many people are aware of the importance of animal-source foods in Timor-Leste, and in 372 particular the importance of fish [3], many women and children suffer from in adequate diets [45] . 373

This study, utilizing robust evidence from interventions both alone and in combination, will 374 generate important knowledge on household barriers to consuming more fish. In particular, our 375 study will determine if additional encouragement to consume fish, when combined with the 376 resources to purchase fish (via a Village Savings and Loan Association) is effective in increasing 377 fish consumption, even without increased supply. As previously stated, there are other factors 378 that can impact fish consumption, including seasonality, fish price, and decision-making power 379 over household income expenditure and cultural practices. Alternatively, we may see that 380

increased supply alone, in combination is additional household spending money is enough to 381 increase fish consumption, even absent of an SBC campaign. These sets of interventions have 382 not been tested in this context before, and the evidence will thus provide important information to 383 policy-makers and stakeholders working to improved nutrition in small island developing states. 384

There are two limitations of this trial. First, the dietary data on fish consumption is based on recall 386

and self-reporting, and therefore risks response bias. It is expected that response bias would 387 favour over-reporting of consumption, which may suggest lower effectiveness of the interventions. 388

Sharing of foods with family members would likely be under-reported, again leading to 389 underestimation of effectiveness. Collection of data on potential confounders / modifiers of effect. 390

While gender and power dynamics are recognised as important influencers of dietary choice 391 and diversity, the intersection of these factors with increased supply and knowledge of nutrition 392

benefits of fish will not be explored in this study. However, VSLAs are composed of both men 393

and women members and SBC activities will involve all members and encourage collaborative 394 food purchasing decision-making. 395 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted August 12, 2021. ; https://doi.org/10.1101/2021.08.10.21261568 doi: medRxiv preprint

The price of fish is a potential confounder in the study; however, we will collect market price and 396

spending data on fish and other animal source foods. The market price variation of fish across 397 sites (e.g. based on the geographical distance between producer and consumer) may affect 398 access and thereby influence dietary preference. Information on prices and spending at each of 399 the sites will be routinely collected as part of this study, and if variation occurs, will be included as 400 a covariate in our adjusted models. In the same way, variable fish quality and safety between 401 sites (fish deteriorating further with longer exposure to heat in reaching further sites) may affect 402 consumer dietary preference. In baseline and endline surveys, we will ask if fish quality is a 403

concern or constraint to purchasing behaviour at each site. Travel time will be established to each 404 site under normal conditions, and temperature loggers will also be used to track the temperature 405 inside coolers carried by traders from producer to consumer sites. 406

From a programmatic point of view, if FADs and SBC prove successful at increasing rates and 408

volume of fish consumption, not only would it provide an acceptable and scalable program, but it 409 might also highlight where other technological innovations could be paired with SBC to drive 410 sustainable market effects without the need for costly incentives. 411

This study will contribute broadly to the body of knowledge on fish consumption patterns and 412

dietary choice by the rural poor in Timor-Leste, and due to close collaboration by WorldFish and 413

Mercy Corps with the Ministry of Agriculture and Fisheries and the FAO, our findings will be used 414

to inform decision making around new interventions, policies and investments in Timor-Leste. The 415 effects of FADs on the supply of fish to inland areas will further highlight possible opportunities 416 for using this technology to achieve national development priority objectives around nutrition and 417 rural development. Furthermore, new knowledge generated about the effects of SBC on dietary 418

choice with specific relevance to fish are of broad interest to those studying food systems, and 419 the dynamics of poverty in small-island developing states and beyond. 420

Peskiza Ai-Han No Nutrisaun 2020. Ministry of Health, GOTL

Malnutrition in Timor-Leste:A review of the burden, 423 drivers, and potential response

Fish and animal protein consumption and availability in Timor-Leste. 426 Regional Fisheries Livelihoods Programme for South and Southeast

The State of World Fisheries and Aquaculture 2020

Malaysia: WorldFish

Aggregating Devices Show Positive Outcomes for Sustainable Fisheries Development in Timor-437

The contribution of nearshore 439 fish aggregating devices (FADs) to food security and livelihoods in Solomon Islands

The benefits of fish aggregating devices in the Pacific

Positive results of a FAD monitoring programme in Yap

FAD: Fish Aggregating Device or Fish Attracting Device? A new 446 analysis of yellowfin tuna movements around floating objects

Tuna Purse-Seine Fishing with Fish-Aggregating Devices (FAD): Models of 448 Tuna FAD Interactions

Biodiversity enhances reef fish 450 biomass and resistance to climate change

Inland fisheries critical for 452 the diet quality of young children in sub-Saharan Africa

Using formative research to 454 design context-specific animal source food and multiple micronutrient powder interventions to 455 improve the consumption of micronutrients by infants and young children in Tanzania

Gender Equality and Social Inclusion Analysis. TOMAK

Factors influencing animal-460 source food consumption in Timor-Leste

A behaviour change 462 intervention with lipid-based nutrient supplements had little impact on young child feeding 463 indicators in rural Kenya

The influence of persuasive messages on healthy eating habits: a test of the Theory of 465

Reasoned Action when attitudes and subjective norm are targeted for change

Lessons from using 468 cluster-randomized evaluations to build evidence on large-scale nutrition behavior change 469 interventions

Using ethnography in implementation research to improve 471 nutrition interventions in populations

The Impact of Savings Groups on Food Security & 474 Nutrition: What does the evidence show us? Australian Aid, TOMAK, Mercy Corps

PeskAAS: A near-real-time, open-source monitoring and 481 analytics system for small-scale fisheries

Minimum Dietary Diversity for Women: A Guide for Measurement. FAO and FHI 360

Global Nutrition Monitoring Framework: operational guidance for tracking progress in 485 meeting targets for 2025. Geneva: World Health Organization

Lee DS. Training, wages, and sample selection: Estimating sharp bounds on treatment effects

Field Experiments: Design, Analysis, and Interpretation

Behavioral Responses and the Impact of New 491

Agricultural Technologies: Evidence from a Double-blind Field Experiment in Tanzania

One Size Fits All? Experimental Evidence on the 494

Digital Delivery of Personalized Extension Advice in Nigeria

The Romano-Wolf Multiple Hypothesis Correction in Stata

Multiple Inference and Gender Differences in the Effects of Early Intervention: A 499 Reevaluation of the Abecedarian, Perry Preschool, and Early Training Projects

Identifying and spurring high-growth entrepreneurship: Experimental evidence from a 502 business plan competition. The World Bank

Understanding and Misunderstanding Randomized Controlled Trials

How to run an experimental auction: a review of 506 recent advances

Controlling Bias in Observational Studies: A Review. Sankhyā: The Indian 508

Global Hunger Index: 510 One Decade to Zero Hunger: Linking Health and Sustainable Food Systems

Livelihood diversity and dynamism in Timor-Leste; 516 insights for coastal resource governance and livelihood development

Consultancy to Verify National RFLP Work Plan Activities, Collaborators and 519 Indicators -Final Report on Timor-Leste. Regional Fisheries Livelihoods Programme for South and 520 Southeast Asia

Anchored Fish Aggregating Devices in Hawaiian Waters. 522 Mar Fish Rev

Fish aggregation device (FAD) research: Gaps in current knowledge and 524 future directions for ecological studies

The use of anchored FADs in the area served by the Secretariat of the 526

Pacific Community (SPC): regional synthesis. Pêche thonière et dispositifs de concentration de 527 poissons

Not just a passing FAD: Insights from the use of artisanal fish 530 aggregating devices for food security in Kiribati

Maternal diets matter for children's dietary quality: 532 Seasonal dietary diversity and animal-source foods consumption in rural Timor-Leste. Matern Child 533 Nutr